Logs and status

hearthwork logs              # the host's log: requests, queue, start and stop, repairs
hearthwork logs --server     # llama-server's own output
hearthwork logs -f           # follow the log as it grows
hearthwork logs -n 200       # show more lines

Use the host log to see how requests were queued and when tool calls were repaired. Use --server when the model failed to load or crashed: it holds llama.cpp’s own messages, such as out-of-memory errors.

To run the host in a terminal and watch its log directly, use hearthwork serve (Ctrl+C stops it).

Status

hearthwork status [--json] shows the host, model, context, slots, requests per client, the budget, switching, idle and recovery state, the sharing route and the active profile. After a crash it shows the last failure, the exit code and the last 20 lines of the server log.

Where the files are

The host log, server.log, logs/requests.jsonl and Doctor reports are all in your data folder.

Debug dumps

Set HEARTHWORK_RELAY_DUMP=<folder> to write every request before and after the relay’s cleanup. Set HEARTHWORK_CAPTURE=1 to also record full request and response bodies. Both contain your code and conversations. Use them only locally and delete the files afterwards.